The architecture inspiring the Neural Network is U-Net: Convolutional Networks for Biomedical Image Segmentation
We trained the model with 40 epochs and a batchsize of 5.
Due to sensitive data, we are not permitted to share the images used for training. Therefore a Jupyter Notebook is not added to this Github. All scripts and code used for "SEGMENTING CAR PARTS USING INTERNAL DATASET" is included in this Github. Due to size limitations the final trained model has been uploaded as zip file.
If it is desired, we can try and reach out to Deloitte and ask if we can upload one image, so a test recreation would be possible.